| 1 | |
| 2 | * linux/fs/proc/root.c
|
| 3 | *
|
| 4 | * Copyright (C) 1991, 1992 Linus Torvalds
|
| 5 | *
|
| 6 | * proc root directory handling functions
|
| 7 |
|
| 8 |
|
| 9 | #include <asm/uaccess.h>
|
| 10 |
|
| 11 | #include <linux/errno.h>
|
| 12 | #include <linux/time.h>
|
| 13 | #include <linux/proc_fs.h>
|
| 14 | #include <linux/stat.h>
|
| 15 | #include <linux/init.h>
|
| 16 | #include <linux/sched.h>
|
| 17 | #include <linux/module.h>
|
| 18 | #include <linux/bitops.h>
|
| 19 | #include <linux/user_namespace.h>
|
| 20 | #include <linux/mount.h>
|
| 21 | #include <linux/pid_namespace.h>
|
| 22 | #include <linux/parser.h>
|
| 23 |
|
| 24 | #include "internal.h"
|
| 25 |
|
| 26 | static int proc_test_super(struct super_block *sb, void *data)
|
| 27 | {
|
| 28 | return sb->s_fs_info == data;
|
| 29 | }
|
| 30 |
|
| 31 | static int proc_set_super(struct super_block *sb, void *data)
|
| 32 | {
|
| 33 | int err = set_anon_super(sb, NULL);
|
| 34 | if (!err) {
|
| 35 | struct pid_namespace *ns = (struct pid_namespace *)data;
|
| 36 | sb->s_fs_info = get_pid_ns(ns);
|
| 37 | }
|
| 38 | return err;
|
| 39 | }
|
| 40 |
|
| 41 | enum {
|
| 42 | Opt_gid, Opt_hidepid, Opt_err,
|
| 43 | };
|
| 44 |
|
| 45 | static const match_table_t tokens = {
|
| 46 | {Opt_hidepid, "hidepid=%u"},
|
| 47 | {Opt_gid, "gid=%u"},
|
| 48 | {Opt_err, NULL},
|
| 49 | };
|
| 50 |
|
| 51 | static int proc_parse_options(char *options, struct pid_namespace *pid)
|
| 52 | {
|
| 53 | char *p;
|
| 54 | substring_t args[MAX_OPT_ARGS];
|
| 55 | int option;
|
| 56 |
|
| 57 | if (!options)
|
| 58 | return 1;
|
| 59 |
|
| 60 | while ((p = strsep(&options, ",")) != NULL) {
|
| 61 | int token;
|
| 62 | if (!*p)
|
| 63 | continue;
|
| 64 |
|
| 65 | args[0].to = args[0].from = NULL;
|
| 66 | token = match_token(p, tokens, args);
|
| 67 | switch (token) {
|
| 68 | case Opt_gid:
|
| 69 | if (match_int(&args[0], &option))
|
| 70 | return 0;
|
| 71 | pid->pid_gid = make_kgid(current_user_ns(), option);
|
| 72 | break;
|
| 73 | case Opt_hidepid:
|
| 74 | if (match_int(&args[0], &option))
|
| 75 | return 0;
|
| 76 | if (option < 0 || option > 2) {
|
| 77 | pr_err("proc: hidepid value must be between 0 and 2.\n");
|
| 78 | return 0;
|
| 79 | }
|
| 80 | pid->hide_pid = option;
|
| 81 | break;
|
| 82 | default:
|
| 83 | pr_err("proc: unrecognized mount option \"%s\" "
|
| 84 | "or missing value\n", p);
|
| 85 | return 0;
|
| 86 | }
|
| 87 | }
|
| 88 |
|
| 89 | return 1;
|
| 90 | }
|
| 91 |
|
| 92 | int proc_remount(struct super_block *sb, int *flags, char *data)
|
| 93 | {
|
| 94 | struct pid_namespace *pid = sb->s_fs_info;
|
| 95 | return !proc_parse_options(data, pid);
|
| 96 | }
|
| 97 |
|
| 98 | static struct dentry *proc_mount(struct file_system_type *fs_type,
|
| 99 | int flags, const char *dev_name, void *data)
|
| 100 | {
|
| 101 | int err;
|
| 102 | struct super_block *sb;
|
| 103 | struct pid_namespace *ns;
|
| 104 | char *options;
|
| 105 |
|
| 106 | if (flags & MS_KERNMOUNT) {
|
| 107 | ns = (struct pid_namespace *)data;
|
| 108 | options = NULL;
|
| 109 | } else {
|
| 110 | ns = task_active_pid_ns(current);
|
| 111 | options = data;
|
| 112 |
|
| 113 | if (!capable(CAP_SYS_ADMIN) && !fs_fully_visible(fs_type))
|
| 114 | return ERR_PTR(-EPERM);
|
| 115 |
|
| 116 |
|
| 117 | if (!ns_capable(ns->user_ns, CAP_SYS_ADMIN))
|
| 118 | return ERR_PTR(-EPERM);
|
| 119 | }
|
| 120 |
|
| 121 | sb = sget(fs_type, proc_test_super, proc_set_super, flags, ns);
|
| 122 | if (IS_ERR(sb))
|
| 123 | return ERR_CAST(sb);
|
| 124 |
|
| 125 | if (!proc_parse_options(options, ns)) {
|
| 126 | deactivate_locked_super(sb);
|
| 127 | return ERR_PTR(-EINVAL);
|
| 128 | }
|
| 129 |
|
| 130 | if (!sb->s_root) {
|
| 131 | err = proc_fill_super(sb);
|
| 132 | if (err) {
|
| 133 | deactivate_locked_super(sb);
|
| 134 | return ERR_PTR(err);
|
| 135 | }
|
| 136 |
|
| 137 | sb->s_flags |= MS_ACTIVE;
|
| 138 | }
|
| 139 |
|
| 140 | return dget(sb->s_root);
|
| 141 | }
|
| 142 |
|
| 143 | static void proc_kill_sb(struct super_block *sb)
|
| 144 | {
|
| 145 | struct pid_namespace *ns;
|
| 146 |
|
| 147 | ns = (struct pid_namespace *)sb->s_fs_info;
|
| 148 | if (ns->proc_self)
|
| 149 | dput(ns->proc_self);
|
| 150 | kill_anon_super(sb);
|
| 151 | put_pid_ns(ns);
|
| 152 | }
|
| 153 |
|
| 154 | static struct file_system_type proc_fs_type = {
|
| 155 | .name = "proc",
|
| 156 | .mount = proc_mount,
|
| 157 | .kill_sb = proc_kill_sb,
|
| 158 | .fs_flags = FS_USERNS_MOUNT,
|
| 159 | };
|
| 160 |
|
| 161 | void __init proc_root_init(void)
|
| 162 | {
|
| 163 | int err;
|
| 164 |
|
| 165 | proc_init_inodecache();
|
| 166 | err = register_filesystem(&proc_fs_type);
|
| 167 | if (err)
|
| 168 | return;
|
| 169 |
|
| 170 | proc_self_init();
|
| 171 | proc_symlink("mounts", NULL, "self/mounts");
|
| 172 |
|
| 173 | proc_net_init();
|
| 174 |
|
| 175 | #ifdef CONFIG_SYSVIPC
|
| 176 | proc_mkdir("sysvipc", NULL);
|
| 177 | #endif
|
| 178 | proc_mkdir("fs", NULL);
|
| 179 | proc_mkdir("driver", NULL);
|
| 180 | proc_mkdir("fs/nfsd", NULL);
|
| 181 | #if defined(CONFIG_SUN_OPENPROMFS) || defined(CONFIG_SUN_OPENPROMFS_MODULE)
|
| 182 |
|
| 183 | proc_mkdir("openprom", NULL);
|
| 184 | #endif
|
| 185 | proc_tty_init();
|
| 186 | #ifdef CONFIG_PROC_DEVICETREE
|
| 187 | proc_device_tree_init();
|
| 188 | #endif
|
| 189 | proc_mkdir("bus", NULL);
|
| 190 | proc_sys_init();
|
| 191 | }
|
| 192 |
|
| 193 | static int proc_root_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat
|
| 194 | )
|
| 195 | {
|
| 196 | generic_fillattr(dentry->d_inode, stat);
|
| 197 | stat->nlink = proc_root.nlink + nr_processes();
|
| 198 | return 0;
|
| 199 | }
|
| 200 |
|
| 201 | static struct dentry *proc_root_lookup(struct inode * dir, struct dentry * dentry, unsigned int flags)
|
| 202 | {
|
| 203 | if (!proc_lookup(dir, dentry, flags))
|
| 204 | return NULL;
|
| 205 |
|
| 206 | return proc_pid_lookup(dir, dentry, flags);
|
| 207 | }
|
| 208 |
|
| 209 | static int proc_root_readdir(struct file *file, struct dir_context *ctx)
|
| 210 | {
|
| 211 | if (ctx->pos < FIRST_PROCESS_ENTRY) {
|
| 212 | int error = proc_readdir(file, ctx);
|
| 213 | if (unlikely(error <= 0))
|
| 214 | return error;
|
| 215 | ctx->pos = FIRST_PROCESS_ENTRY;
|
| 216 | }
|
| 217 |
|
| 218 | return proc_pid_readdir(file, ctx);
|
| 219 | }
|
| 220 |
|
| 221 | |
| 222 | * The root /proc directory is special, as it has the
|
| 223 | * <pid> directories. Thus we don't use the generic
|
| 224 | * directory handling functions for that..
|
| 225 |
|
| 226 | static const struct file_operations proc_root_operations = {
|
| 227 | .read = generic_read_dir,
|
| 228 | .iterate = proc_root_readdir,
|
| 229 | .llseek = default_llseek,
|
| 230 | };
|
| 231 |
|
| 232 | |
| 233 | * proc root can do almost nothing..
|
| 234 |
|
| 235 | static const struct inode_operations proc_root_inode_operations = {
|
| 236 | .lookup = proc_root_lookup,
|
| 237 | .getattr = proc_root_getattr,
|
| 238 | };
|
| 239 |
|
| 240 | |
| 241 | * This is the root "inode" in the /proc tree..
|
| 242 |
|
| 243 | struct proc_dir_entry proc_root = {
|
| 244 | .low_ino = PROC_ROOT_INO,
|
| 245 | .namelen = 5,
|
| 246 | .mode = S_IFDIR | S_IRUGO | S_IXUGO,
|
| 247 | .nlink = 2,
|
| 248 | .count = ATOMIC_INIT(1),
|
| 249 | .proc_iops = &proc_root_inode_operations,
|
| 250 | .proc_fops = &proc_root_operations,
|
| 251 | .parent = &proc_root,
|
| 252 | .name = "/proc",
|
| 253 | };
|
| 254 |
|
| 255 | int pid_ns_prepare_proc(struct pid_namespace *ns)
|
| 256 | {
|
| 257 | struct vfsmount *mnt;
|
| 258 |
|
| 259 | mnt = kern_mount_data(&proc_fs_type, ns);
|
| 260 | if (IS_ERR(mnt))
|
| 261 | return PTR_ERR(mnt);
|
| 262 |
|
| 263 | ns->proc_mnt = mnt;
|
| 264 | return 0;
|
| 265 | }
|
| 266 |
|
| 267 | void pid_ns_release_proc(struct pid_namespace *ns)
|
| 268 | {
|
| 269 | kern_unmount(ns->proc_mnt);
|
| 270 | }
|